News, Fast

Atom Brief

Jul 22, 2026 · Archive
Tech

Moonshot's Kimi K3 Beats Every AI Model But Anthropic's Fable 5

On Artificial Analysis' AA-Briefcase test, Kimi K3 scored 1543 Elo to Fable 5's 1574, topping GPT-5.6 Sol and two Claude models.

  • Kimi K3 scored 1543 Elo on Artificial Analysis' AA-Briefcase agentic knowledge benchmark, published July 21, 2026.
  • Claude Fable 5 led at 1574 Elo, the only model to outscore Moonshot's Kimi K3.
  • Kimi K3 beat GPT-5.6 Sol, Claude Sonnet 5, and Claude Opus 4.8 on the same test.

Why it matters: The compute gap between America's and China's frontier labs is now single digits, not generations.

Artificial Analysis (AA-Briefcase article) ↗ · Jul 22, 20267/22/26